Kink related excess noise in PD SOI MOSFETs

Due to the floating body, SOI MOSFETs exhibit a low frequency kink-related noise overshoot, which has a Lorentzian spectrum: a flat low-frequency plateau, with constant amplitude, followed by a 1/f² roll-off. Therefore, the kink-related excess noise is also called Lorentzian-like noise. Several mechanisms have been proposed to explain this excess noise, such as trap-assisted generation- recombination (G-R) noise caused by the defects in body region [13], G-R noise induced by back interface state caused by the defects at bottom Si-SiO2 interface [14], and shot noise amplified by the floating body effect [15]. In early years, owing to the worse process, the G-R noise is the main source of the excess noise. At present, the technology of the process is quite advanced, thus the defects in Si film and bottom Si-SiO2

interface are fewer and fewer. Therefore, shot noises from source to body and drain to body

become the major sources. The shot noise sources from drain to body and source to body result from the drain-body junction leakage (I_L) and body-source diode current (I_SB) respectively (see Fig. 2-2). Although both are shot noises generated by flow of carriers surmounting energy barriers, they are independent noise sources. However, these two noises are basically the same, because in SOI IL is roughly equal to ISB as a result of current balancing in the body of SOI device. Here we defined the region, which the drain biasis smaller than the dc kink onset voltage, as “pre-kink region”. On the contrary, when the drain biasis larger than the dc kink onset voltage, the region is defined as “post-kink region” (see Fig. 2-3). In the pre-kink region, where the drain voltage is small, impact ionization can be neglected; therefore I_L is dominated by junction thermal generation current (I_G). In the post-kink region, where the drain voltage is large enough to lead to impact ionization, the IL is dominated by impact ionization current (I_ii).

The shot noise originates in the IL and ISB are small in magnitude compared with flicker noise, but the high impedance of source-body junction caused by floating body effect significantly amplifies their magnitude and gives rise to the excess low-frequency noise in PD SOI MOSFET. From the shot noise model proposed by Jin [16], the shot noises can be express as (2qISB+2qIL) which are a white noise. The source-body junction impedance (Zbody) is the parallel combination of r_SB and C_BB. These current fluctuations perturb the body voltage through the square of the body impedance with Lorentzian-like shape similar to a RC filter shown in Fig 2-4, and no longer be a white noise [17]. After that, these body voltage fluctuations transferred through the body effect to gate voltage power spectral density (V²/Hz), as followed:

VG excess shot Body SB L

BS S

source-body junction resistance, C_BB is the body-ground capacitance, and f₀ =qI_L /(2π nkTCBB

). Therefore, the corner frequency f₀ is proportional to I_L and the plateau is proportional to 1/

IL.

As the operation mode shifts from PD toward more FD operation, source-body junction diode saturation current, I_R (where source-body junction diode current I_SB=I_R⋅(exp(V_SB/nV_T)-1)) exponentially increases due to the reduced source-body junction barrier, as follows:

(2-3) ( ) ( ) ^qVBS/^kT

R R

I FD =I PD ×e ^′

where V'BS is the reduced junction barrier potential. Exponential increase of IR in FD SOI devices results in a significant decrease of source-body junction impedance Z_body (i.e. shot noise amplified gain). Therefore, the floating body effect is suppressed by an increase of IR, and I_R has to be included in the excess noise model with a modification of r'_SB=nV_T/(I_SB+I_R) as operation toward more FD, I_R will increase and thus lead to the increase of f'₀ and reduction of the plateau. If the magnitude of plateau is smaller than 1/f noise, then we can observe only 1/f noise, i.e. the shot noise is suppressed.

The other way to reduce excess noise is using body contact. A good body contact provides another low-resistance discharge path for body charges, and the impedance of current fluctuation can drastically reduce (i.e. shot noise amplified gain can reduce), as follows:

B body body

R &Z <Z (2-5)

where RB is the body path resistance. The criterion for low frequency noise optimization is to suppress the Lorentzian-like noise overshoot making it less than the underlying 1/f noise by decreasing the body transfer function at the desired biases.

1/f ²

Fig. 2-1 Schematic plot of low frequency noise components

BOX

Fig. 2-2 Noise sources in a SOI MOSFET. :1/f noise : shot noise due to body-source diode current

: shot noise due to drain-body junction leakage

2

i

i

i

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0.0

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

VG=1.25V

VG=0.5V VG=0.75V VG=1.0V

Post-kink

I D (mA)

V_D (V)

DC kink onset voltage

Pre-kink

PD SOI W/L=10/0.2

Fig. 2-3 ID-VD curves in floating body SOI MOSFET

S

2

v

r

C

2qI_L 2qISB

D

2

i

C

2

v

r

Fig. 2-4 Noise small-signal equivalent circuits for the floating body SOI MOSFET

Chapter 3

Characterization of Low-Frequency Noise in PD SOI MOSFETs

In this chapter, we describe the measurement methods for I-V and low-frequency noise characterization firstly. Then we discuss the noise characteristics of PD SOI MOSFETs with floating-body and source-to-body-connected structures. Because of the continuous scaling-down of devices for better performance, it is important to realize and model the low-frequency noise in SOI MOSFETs after channel shrinking for device process diagnosis and analog circuit simulation. Hence the influence of channel length on the low-frequency noise of PD SOI MOSFETs also has been investigated in this chapter.

3.1 Devices under Test and Measurement Techniques

3.1.1 Devices under Test

The n-channel MOSFETs were fabricated on Separation by Implantation of Oxygen (SIMOX) substrates with 190nm thick Si active layers, 150nm thick buried oxides, and 1.6nm nitride gate oxides. The floating-body and source-to-body-connected transistors with oxide/SiN spacers and As⁺-implanted source/drain junctions are partially depleted. After CoSi2 salicidation, the devices were metalized using a typical backend flow.

3.1.2 I-V Measurement

The DC characteristics of PD SOI MOSFETs were measured using HP4145 or HP4156.

From the ID-VG curves, we extract the threshold voltage (VTH), transconductance (gm=dID

/dVG ) and body factor (α=dVTH /dVBS ) at different channel lengths. We also observed the kink effect from the ID-VG and ID-VD characteristics.

3.1.3 Noise Measurement

The HP4145, HP35670A Dynamic Signal Analyzer, and BTA9812 Noise Analyzer were employed to measure the low frequency noise at different channel lengths, where the measurement system was fully automated using a personal computer via HP-IB. The measurement frequency range is from 1Hz to 10 kHz, and the output drain current noise power spectrum would be extracted. We divide the output drain current noise power spectrum by the square of the transconductance to obtain the input-referred gate voltage noise power spectrum (i.e. SVG=SID /gm²).

The devices operate in linear operation are corresponding to a drain voltage VDS = 0.1V.

As the drain bias increases to about 0.4V, the devices enter the saturation region, both the data obtained before the kink occurring (VDS = 0.8V) and after the kink occurring (VDS = 1.0V) are shown.